
\begin{table}[h]
\begin{center}
\begin{tabular}{l D{.}{.}{6.5} D{.}{.}{6.6} D{.}{.}{6.5}}
\toprule
 & \multicolumn{1}{c}{Model 1} & \multicolumn{1}{c}{Model 2} & \multicolumn{1}{c}{Model 3} \\
\midrule
UPR Shaming (t-1)         & 0.0004                  & 0.0006^{**}             & 0.0006^{*}              \\
                          & (0.0003)                & (0.0003)                & (0.0003)                \\
\midrule
Individual-level Controls & \multicolumn{1}{c}{No}  & \multicolumn{1}{c}{Yes} & \multicolumn{1}{c}{Yes} \\
Physical Integrity (t-1)  & \multicolumn{1}{c}{No}  & \multicolumn{1}{c}{No}  & \multicolumn{1}{c}{Yes} \\
Imputed Capacity (t-1)    & \multicolumn{1}{c}{No}  & \multicolumn{1}{c}{No}  & \multicolumn{1}{c}{Yes} \\
Year FEs                  & \multicolumn{1}{c}{Yes} & \multicolumn{1}{c}{Yes} & \multicolumn{1}{c}{Yes} \\
Country FEs               & \multicolumn{1}{c}{Yes} & \multicolumn{1}{c}{Yes} & \multicolumn{1}{c}{Yes} \\
Num. obs.                 & 120277                  & 120277                  & 120277                  \\
Num. groups: year         & 12                      & 12                      & 12                      \\
Num. groups: country.x    & 45                      & 45                      & 45                      \\
Deviance                  & 27267.5466              & 25257.1062              & 25242.9471              \\
Log Likelihood            & -81415.1107             & -76809.1223             & -76775.3994             \\
Pseudo R$^2$              & 0.0455                  & 0.0994                  & 0.0997                  \\
\bottomrule
\multicolumn{4}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.05$; $^{*}p<0.1$}}
\end{tabular}
\caption{The effect of UPR shaming on incumbent voting (CSES). Logistic Regression Models. Robust standard errors are clustered at the country-year level. Missing data for control variables is imputed (average value by country-year).}
\label{tab:H1_GLM}
\end{center}
\end{table}
